Understand Kubernetes Management

In the GlobalAI Infrastructure module, Kubernetes management goes beyond defining standalone manifests. The platform provides a structured system for creating reusable templates, deploying them across clusters, and managing their full lifecycle.

This framework uses three core capabilities:

• Application templates: Versioned blueprints that define how your application deploys.
• Cluster groups: Logical collections of clusters that support automatic target selection based on deployment rules.
• Deployment lifecycle controls: Automated resource-management features that help reduce costs in non-production environments.

Application templates

Application templates act as reusable blueprints for deploying workloads. They enable consistent, repeatable deployments across clusters.

Key characteristics include:

• Standardization: Templates typically use Helm charts and you can upload them as .zip or .tar.gz packages. With them, you can define parameters, namespaces, and configuration values in a structured way.

• Versioning: Each template can have multiple versions. Once a version has at least one deployment, it becomes read-only to ensure consistency.

• Updates & revisions: When you update a deployment to a newer version, the platform creates a revision. You can review previous revisions and revert if needed.

Cluster groups

As your infrastructure grows, manually selecting the correct cluster for each deployment becomes inefficient. Cluster Groups simplify this process by letting you treat clusters as a single logical destination.

Each group uses a Deployment Rule to determine how clusters are selected:

Rule Type	Description
Random	The system chooses a cluster at random from the group. Useful for stateless or interchangeable workloads.
Round Robin	The system selects clusters sequentially to balance deployments across the group.

When you deploy a template version associated with a Cluster Group:

• The platform automatically selects the target cluster based on the rule.
• You may still override the choice and select a different cluster manually.

Geo-specific deployments

Use Cluster Groups to enforce regional placement, for example, a “Europe” group ensures deployments remain within EU-approved clusters.

Deployment lifecycle controls

GlobalAI includes lifecycle settings that help optimize infrastructure costs, especially for development, demo, or temporary environments.

Each deployment can define:

• Default Time (how long the deployment stays active)
• Additional Time / Grace Period
• Time values use the ISO 8601 duration format (for example, P1D = 1 day)

Deployment lifecycle options

You can configure each deployment with a lifecycle option that determines how GlobalAI manages its compute resources and data after the defined time window expires.

Option	Resource Impact	Data Persistence	Best Use Case
Manual	Resources remain active until you remove them.	Full persistence.	Production workloads or long-running environments that must remain online.
Stop	The platform halts workload resources (CPU and RAM) while preserving storage.	The platform retains secrets, configs, and Persistent Volumes.	Development or testing environments you want to pause without losing data.
Terminate	The platform removes all resources when the configured time expires.	No persistence.	Temporary demos, previews, or POCs that require full cleanup after use.

Related articles

Now that you understand the core concepts behind Kubernetes management, explore the next steps:

Configure Clusters

Connect your Kubernetes clusters and configure monitoring.

Manage Cluster Groups

Set up routing rules and organize your clusters.

Deploy Applications

Create deployments using your templates.